A Wireless Temperature Sensor Powered by A

Piezoelectric Resonant Energy Harvesting System

Peng Wang1*, Robert Gray1,2, Zenghui Wang1, Philip X.-L. Feng1*

1
Electrical Engineering and Computer Science, Case Western Reserve University, Cleveland, OH 44106, USA
2
Hawken High School, Gates Mills, OH 44040, USA
*
Email: [email protected], [email protected]

AbstractWe report on experimental demonstration of a dominant power consumption of a wireless temperature sensor
wireless temperature sensor node (WTSN) powered by a node (WTSN) comes from RF communication, while the
piezoelectric resonant energy harvesting system. The energy averaged power for temperature reading is usually only
harvesting circuit stores the energy generated by a piezoelectric hundreds of nW. Fig. 1 shows that with the sensor power
resonant transducer into a capacitor, and uses the stored energy consumption level under the current technology, battery
to power the temperature sensor and its associated signal replacement is inevitable on a yearly basis, if not monthly.
processing circuits for wireless signal transmission. The main Hence to extend the maintenance-free working span of a
functions of the harvesting system are implemented by discrete wireless sensor, self-powering solutions such as vibration
components together with a power management application
energy harvesting are required.
specific integrated circuit (ASIC). The resulting WTSN
transmits measured temperature data over a distance of 10m, in 1k Minute
real time, and consumes ~4W to ~13W power (for Alkaline AA Battery
100
transmission intervals from 10min to 10s, respectively), which is 20mm Li Coin Battery
supplied entirely by the piezoelectric resonant energy harvester. 10 Hour
Usually 3-4 Years
1
Power Budget (W)

Day Limited by Battery

Keywordswireless temperature sensor; energy harvesting; 100m Lifetime
piezoelectric resonator; transducer; lead zirconate titantate (PZT) Month
10m
Year
I. INTRODUCTION 1m
100 Typical WTSN Decade
Wireless temperature sensor networks are key to energy-
Power Range
efficient smart buildings for more sustainable and resource- 10
conserving development [1]. However, battery-powered large- 1
Energy
scale temperature sensor networks have high maintenance costs Harvesting
100n
due to required periodical replacements of batteries, which 100 101 102 103 104 105 106 107 108 109 1010 1011
limit their real-world applications and deployments. Vibration Lifetime (s)
energy harvesting offers a viable alternative to battery-powered Fig. 1: Power budget analysis for typical commercial wireless temperature
wireless temperature sensor networks, as it enables self- sensors, with examples shown for battery-powered options.
powered sensor nodes which shall no longer require battery
replacements, and thus can lead to maintenance-free systems. Here we report on an experimental demonstration of using
a piezoelectric resonant energy harvesting system to power a
Wireless sensor networks based on 802.15.4/ZigBee are WTSN. We customize a PZT transducer, which resonates at
promising for smart building applications, and the wireless the same frequency (120Hz) with the vibration source (a small
protocol facilitates flexibility, interoperability, and large service pump, or devices alike in a regular ventilation service
number of sensor nodes. A typical 802.15.4/ZigBee sensor room). We use discrete components and an ASIC prototype
network consists of front-end sensor devices (wireless sensor for the energy conversion and storage functions. With this
nodes), repeaters (wireless relays), and coordinators (wireless energy harvesting system, we successfully demonstrate a
hubs). The repeaters and coordinators are usually powered WTSN, which is capable of transmitting measured temperature
through mains and have high operating power (thus high over a distance of up to 10m, and consumes ~13W
receiver sensitivity), which can in turn reduce the power (transmitting at 10s interval) or 4W (at 10min interval)
consumption and duty cycle on the front-end devices power, all supplied by the vibration energy harvester.
(transmitters). The front-end sensor devices require ultra-low
power budget design, which, together with vibration energy II. SYSTEM DESIGN
harvesting, can realize self-powered sensor nodes. A. System Block Diagram
Fig. 1 illustrates the power budget analysis for a wireless Fig. 2 shows the block diagram of the WTSN system. It
sensor node. The power consumption range of a typical includes a PZT resonant energy harvester, a full-wave-bridge
wireless front-end sensor device is ~5W to ~1mW [2-5], rectifier, an ASIC prototype, and an RF module with a
based on transmission intervals (gaps) of 10s to 10min. The microcontroller integrated with a temperature sensor. The full-

978-1-4799-8866-2/15/$31.00 2015 IEEE

wave rectifier is built with Schottky diodes, with a total voltage =0.1227m at 2.445GHz. Fig. 4a shows wireless range d vs.
drop of 0.7V. A 50nF capacitor is used as a filter, and the the transmitter output power PT as calculated using Eq. (1),
harvested energy is stored in a 100F capacitor. with available settings on the CC2500 indicated by square dots.
ASIC The link budget is analyzed by calculating the difference
PZT Power Sleep Control between the received signal power, PR, and the sensitivity of
RF
LDO
the receiver. In a practical design of link budget, additional
Power MCU
Gating Temperature output power has to be added to the output power predicted by
Cfilter Charging
Friis equation. Sufficient link budget can reduce the RX power
Sensor
Cstore eZ430 RF 2500 consumption and data packet loss rate. Fig. 4b shows the link
TI Inc. budget for a typical indoor transmission distance d=10m.
Fig. 2: System diagram of the wireless temperature sensor node (WTSN).
103 (a) (b)
B. PZT Transducer Design 40

Transmission Distance (m)

The PZT transducer takes the form of cantilever-shaped

Link Budget (dBm)

resonator (15mm12mm200m), laser machined to have an 102 20
optimized dimension. The PZT part extends ~2/3 of the total
cantilever length for minimizing the undesirable charge
redistribution effect [6, 7]. Once made, the PZT transducer is 101 0
mounted on the vibration source (a small pump) that has been
characterized using a 3-axis accelerometer (ADXL325, Analog
Devices, Inc.). We then fine-tune the resonance frequency of 100 -20
-60 -50 -40 -30 -20 -10 0 -60 -50 -40 -30 -20 -10 0 10
the PZT cantilever (Fig. 3) to match that of the vibration source Transmitter Output Power (dBm) Transmitter Output Power (dBm)
through adjusting the proof mass, a piece of low-melting- Fig. 4: Wireless transmission analysis results. (a) Calculated wireless
temperature metal attached at the end of the cantilever. Upon transmission range as a function of transmitter output power. (b) Calculated
fine tuning, the optimized PZT transducer can output up to RF link budget for wireless transmission over 10m using CC2500. The dots
represent available configurations in the CC2500.
96W (measured with a 51k load resistor) with a peak
acceleration of 1.5g at the pump surface. E. Temperature Sensor
Output + The temperature sensor used in this study is a silicon (Si)
bandgap temperature sensor integrated in the microcontroller
(MSP430F2274) of the eZ430RF2500 module, and is read by
Proof an on-chip 10-bit ADC converter.
Mass III. EXPERIMENTAL TECHNIQUES
PZT
A. PZT Transducer Characterization
(a) 1cm (b) Cantilever 1cm
We characterize the PZT transducer in both frequency and
time domains. In the frequency domain measurement, the
Fig. 3: PZT transducer mounted on a surface. (a) Front view. (b) Side view.
transmission loss is measured with a network analyzer
C. Wireless Network (Agilent 4395A) as shown in Fig. 5a. In the time domain
Wireless connection (over ZigBee network) is established measurement, we deflect the cantilever by a fixed amount, and
between the WTSN (front-end device) and a receiver (access then suddenly release it. The PZT cantilever then undergoes a
point, AP) connected to a data-taking computer. ring down oscillation which decays over time, and its
voltage output is measured with an oscilloscope (Fig. 5b).
D. Wireless Transmission Range
Network Analyzer
To estimate the wireless transmission range, we calculate Agilent 4395A
Digital Oscilloscope
the RF signal attenuation over distance, wireless link budget,
and power budget on the sensor node.
The wireless signal attenuation can be generally described RF R A B
by the Friis equation: the received power PR at a distance d is:
PR = PTGRGT(/4)2/d n. (1)
Metal Alloy
Here, PT is the originally transmitted power at the source, GR Transmission- Al Film
and GT are the gains of the transmitting (TX) and receiving Reflection Test Kit PZT Initial Deflection and
(RX) antenna respectively, is the EM wavelength, and n is an (a) HP 87512A Brass (b) Abrupt Release
environmental parameter (n=2 in free space; here we use
Fig. 5: Schemes for characterizing the PZT cantilever. (a) Frequency domain
n=2.2, typical value for most indoor settings). In our measurement. We measure the S21 parameter to identify the cantilever
experiment, both the transmitter and receiver (eZ430RF2500 resonance frequency from its frequency-dependent transmission loss. (b)
and CC2500) have an average gain of 1dBi, the receiver Time domain measurement. We measure the ring down oscillations of the
sensitivity is -104dBm at 2.4kBaud 1% packet error rate, and PZT cantilever using an oscilloscope (with an amplifier).
B. PZT Harvester Assembly and Installation remaining part of the WTSN is tested under different
The complete WTSN is shown in Fig. 6a, and the inset temperature settings and wireless environments.
picture shows the prototype ASIC die. The fully assembled The different settings/environments include ambient open
WTSN, once characterized, is mounted on the pump surface space in the lab (room temperature controlled by thermostat),
as shown in Fig. 6b. The PZT transducer energy harvester is on top of a hotplate (HP131725, Thermo Scientific Inc.),
fixed on the pump using 7036 Blanchard Wax (J.H. Young underneath a heat lamp, and inside two ovens with metal
Company) for optimized mechanical coupling. enclosures (model 1410 and 1410MS, VWR International,
eZ430 LLC.). A separate digital temperature sensor (Caliber IV
1cm 1cm
RF2500 Digital Hygromter, Western Humidor, Inc.) is used as a
ASIC reference. During the experiment, the WTSN wirelessly
transmits the temperature reading to the wireless access point
(AP) once every 10s, and the data is logged by the data-taking
PZT
ASIC
eZ430 computer connected to the AP device (similar to the scheme
RF2500
Rotary Pump
during WTSN power consumption characterization, as shown
in Fig. 7). The access point is located 10m away from the
(a) PZT (b)
WTSN for the open space/heat lamp measurements, and 2m
Fig. 6: System assembly. (a) The WTSN. (b) WTSN mounted on the pump. away during the oven/hotplate measurements.
C. WTSN Power Consumption Characterization IV. RESULTS AND DISCUSSIONS
We characterize the transient power consumption of the A. PZT Transducer Characterization
eZ430RF2500 front-end device. As shown in Fig. 7, we power The PZT transducer is characterized using the methods
the WTSN with a battery pack, and connect a 68 resistor in described in Section III-A (and shown in Fig. 5). The results
series. This allows us to estimate the current (and thus power for both measurements are shown in Fig. 8. From the
consumption) in the WTSN by using a digital oscilloscope frequency domain transmission/reflection measurement (Fig.
(TDS1012C-EDU, Tektronix) to monitor the voltage drop 8a), we clearly observe the mechanical resonance of the PZT
across the resistor. transducer exemplified as a sharp heart beat curve feature on
During the characterization, the WSTN is in the normal top of a slowly-varying frequency dependent back ground.
operation mode. It reads the environmental temperature with The data show that our fine tuning results in good frequency
the built-in temperature sensor, and sends the readings matching of the cantilever to the vibration source (120Hz).
wirelessly to the wireless access point (AP, another eZ430 RF The result from the time domain measurement is shown in
2500 unit) connected to the data-taking computer (Fig. 7). Fig. 8b. The data are fit to the ring-down curve of a damped
Digital Oscilloscope harmonic oscillator:
Temperature Reading
f t
a (t ) = a0 + A exp res sin ( 2 f res t + ) , (2)
Q
AP where a(t) is the time-dependent amplitude of the vibration, a0
is the offset of the response, A is the initial amplitude, fres is
the resonance frequency, Q is the quality factor, and is the
initial phase. From the fitting we extract the resonance
RF frequency fres and quality factor Q of the PZT cantilever. The
MCU 68 3.3V results show excellent agreement with the frequency domain
RF Temperature measurement, and again verify the precise tuning of the
MCU Temperature Sensor cantilever frequency.
eZ430 RF 2500 Data
eZ430 RF 2500
Access Point End Device (a) (b) 2 Measurement
Fitting
Output Voltage (V)

Fig. 7: Scheme for measurement of transient power. A small resistor of 68 is -50

connected in series with the power input to the WTSN (the wireless end
S21 (dB)

device), which wirelessly communicates with the wireless access point (AP). 0
D. Wireless Temperature Reading
Once the WTSN is characterized, we demonstrate wireless -60
f = 120.9Hz
temperature sensing powered entirely by vibration energy Q = 24
-2
harvesting. The experimental setup is the same as shown in 110 120 130 0.05 0.10 0.15
Fig. 7, except that the battery pack is replaced by the PZT Frequency (Hz) Time (s)
energy harvesting system, and the current-monitoring resistor
Fig. 8: Characterization of the resonant PZT cantilever energy harvester. (a)
(and the oscilloscope) is removed. While the PZT transducer Frequency domain measurement. (b) Time domain data and fitting. Blue
remains fixed on the pump throughout the experiment, the Circles: Raw data. Red Curve: Fitting of the data according to Eq. (2).
B. Power Consumption of WTSN V. CONCLUSIONS
The power consumption of the WTSN is measured using In summary, we have demonstrated a wireless temperature
the methods described in Section III-C (as shown in Fig. 7). sensor node (WTSN) powered by a PZT resonant transducer
The measured current (and calculated power consumption) that scavenges vibration energy and supplies ~10 to ~100W
during one active transmission event (~30ms long) is shown in to the circuits. The system is fully characterized, and the
Fig. 9a. The peak power is around 5565mW, and lasts about power output of the PZT is sufficient for normal operations of
2ms. Outside this ~30ms active window, the idle power of the WTSN. We have further demonstrated real-time wireless
the WTSN is about 4W, which sets the baseline of the temperature sensing under different experimental settings, with
average power consumption. Based on these measured values, the signal transmission over a distance up to 10m.
we estimate the overall average power consumption of the
(a) 55
WTSN with different transmission intervals (Fig. 9b).
Example values are also shown for 10s (13W) and 10min 2m
50
(4W) transmission intervals, both of which are well within Receiver

Temperature (oC)
the PZT power output of 96W (see Section II-B). 45 42oC
WTSN
1.5 40 36oC

Voltage Drop on Resistor (V)

60 (a)
Transient Power (mW)

35

30 29oC
1.0
40 26oC
25
0 10 20 30 40 50 60 70 80
Time (min)
20 0.5
(b) 100 WTSN
10m
0 0.0 80 Receiver
Temperature ( C)
-20 -10 0 10 20 o
Time (ms) 65oC
104 57oC
60
eZ430 RF 2500 (b) 49oC
Power Requirement (W)

1cm End Device Heat Lamp

3
10 40
10s, 13 W 27oC
102 10min, 4 W 0 10 5 15 20
Limited by Sensor Time (min)
Idle Power Fig. 10: Wireless temperature sensing in real time. (a) Temperature sensing
101 from inside an oven. (b) Temperature sensing underneath a heat lamp. Blue
Circles: Raw data. Red Curves: Smoothed data. Vertical dashed lines indicate
changes in temperature setpoint.
0
10
1000 100 10 1 0.1 ACKNOWLEDGMENT
Transimission Interval (s)
We thank Ran Wei, Aman Nair, Xu-Qian Zheng, Wen H.
Fig. 9: Calibration of power consumption of the WTSN. (a) Measured current Ko, Steve J. A. Majerus, and Christian A. Zorman for technical
in a WTSN during one transmission event (~30ms). (b) Average power support and helpful discussions. We thank support from U.S.
consumption as a function of transmission interval.
Department of Energy (DOE) EERE Grant (DE-EE0006719).
C. Wireless Temperature Reading
REFERENCES
Experiments of vibration-powered wireless temperature
sensing are performed under the settings described in Section [1] G. Lu, D. De, M. Xu, W.-Z. Song, and J. Cao, TelosW: Enabling ultra-
low power wake-on sensor network, in Proc. 7th Int. Conf. on
III-D. Fig. 10 shows examples of the wirelessly transmitted Networked Sensing Systems (INSS), Kassel, Germany, June 15-18, 2010,
temperature data. Fig. 10a shows the measurement performed pp. 211-218. DOI: 10.1109/INSS.2010.5573125.
when the temperature sensor is placed inside an oven, with the [2] E. Casilari, J. M. Cano-Garca, and G. Campos-Garrido, Modeling of
oven temperature going through four setpoints (26C, 29C, current consumption in 802.15. 4/ZigBee sensor motes, Sensors, vol.
36C, and 42C) with 20min duration at each setpoint. During 10, no. 6, pp. 5443-5468, 2010.
this measurement, the wireless receiver (AP) is located at 2m [3] www.ti.com.cn/cn/lit/ds/symlink/tmp102.pdf
away from the WTSN. Fig. 10b shows the measurement when [4] www.ti.com/lit/an/slaa378d/slaa378d.pdf
the WTSN is placed underneath a heat lamp, showing fast [5] www.ti.com/lit/pdf/swra292
response of the temperature sensor when the lamp is turned on [6] M. Stewart, P. M. Weaver, and M. Cain, Charge redistribution in
and off. We also increase the heating duration (on time of the piezoelectric energy harvesters, Appl. Phys. Lett., vol. 100, no. 7,
073901, 2012.
heat lamp) during each heating event, resulting in increasing
[7] R. Wei, P. Wang, W. H. Ko, and P. X.-L. Feng, Characterizing
temperature reading from the WTSN. With the open space piezoelectric cantilevers for vibration energy harvesting under ambient
setting (no metal enclosure), the wireless transmission distance conditions, in Proc. of IEEE EnergyTech 2013, Cleveland, OH, May
is increased to 10m in this measurement. 21-23, 2013, pp. 1-6. DOI: 10.1109/EnergyTech.2013.6645299.